[
https://issues.apache.org/jira/browse/FLINK-20695?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chesnay Schepler updated FLINK-20695:
-------------------------------------
Description:
I used flink 1.11 in standalone cluster mode for batch job. The enviornment was
configured as zookeeper HA mode.
After job was commited, flink runtime created nodes under /flink/default/leader
and /flink/default/leaderlatch with job id. Though jobs were finished, these
nodes were remaining in zookeeper path forever.
After a period of running, more and more jobs had been executed and there were
a greate number of nodes under /flink/default/leader and slowed down the
performance of zookeeper. Why not delete the nodes after job finished? Flink
runtime could get job status by listeners and delete the leader nodes for job
immidiately.
was:
I used flink 1.11 in standalone cluster mode for batch job. The enviornment was
configed as zookeeper HA mode.
After job was commited, flink runtime created nodes under /flink/default/leader
and /flink/default/leaderlatch with job id. Though jobs were finished, these
nodes were remaining in zookeeper path forever.
After a period of running, more and more jobs had been executed and there were
a greate number of nodes under /flink/default/leader and slowed down the
performance of zookeeper. Why not delete the nodes after job finished? Flink
runtime could get job status by listeners and delete the leader nodes for job
immidiately.
> Zookeeper node under leader and leaderlatch is not deleted after job finished
> -----------------------------------------------------------------------------
>
> Key: FLINK-20695
> URL: https://issues.apache.org/jira/browse/FLINK-20695
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Task
> Reporter: lidesheng
> Priority: Critical
>
> I used flink 1.11 in standalone cluster mode for batch job. The enviornment
> was configured as zookeeper HA mode.
> After job was commited, flink runtime created nodes under
> /flink/default/leader and /flink/default/leaderlatch with job id. Though
> jobs were finished, these nodes were remaining in zookeeper path forever.
> After a period of running, more and more jobs had been executed and there
> were a greate number of nodes under /flink/default/leader and slowed down the
> performance of zookeeper. Why not delete the nodes after job finished? Flink
> runtime could get job status by listeners and delete the leader nodes for job
> immidiately.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
